torturing caps

Diode reverse voltage testing isn't destructive.

Anybody who sold untested diodes wouldn't last long.

John

------------

You really have no clue what you are talking about...

$file/f_sec4.pdf

Maximum reverse voltage is destructive BTW, unless you do not actually perform **MAXIMUM** reverse voltage in which case it is not the maximum but your best guess.

If you were as intelligent as you pretend you would at least realize if every component was tested there would be no reason to do any statistical analysis. If you were a real engineering, instead of pretending, you would also realize the cost/yield ratio would be unsustainable and the real issues in quality control are in the front end and not the back end of the production process.

I'm on Larkins side, he is always right... Ever since I received the Mantis he sent me. The thing is wonderful! Mikek

"Friendship" that is purchased.... ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Where's mine?!

Pardon me, but that's blather.

What you do is apply a small constant current and see what the resulting voltage is, essentially measururing the zener point. A couple hundred microamps of reverse current, for a millisecond or so, isn't going to damage any diode. Once you know the breakdown voltage, you can bin the diodes if you want.

Testing machines are a big business. And the only way to get decent process statistics at the low PPM level is to test *every* part. That's also the only way to get repeat customers.

Google will find lots more. Try

Cool, 40,000 diodes per hour:

Rotary turrets rock.

John

You don't have to test the **MAXIMUM** reverse voltage. You only have to test to the spec.

Is that you, AlwaysWrong?

Our experience suggests that most diode makers have at least two different wafer processes, one for low voltage diodes and one for higher ones. The HV diodes behave like PIN diodes (which they pretty much are) and have higher forward drops.

All your arguments are based on the premise that reverse breakdown voltage measurements are destructive. They aren't.

John

This AlwaysWrong thing seems to be getting more popular lately.

John

I was kinda surprised there was no swearing or talk of feces.

Nothing there about negligently shipping out untested devices... in fact most semiconductors are tested at least twice-- once at wafer level and once after packaging. That expensive ABB power stuff looks like its not only tested but 100% burned-in. Worth it for some applications.

Goodness. It's not a "guess". For PIV, the tester attempts to put a constant current through the DUT for a few milliseconds using a source compliant up to hundreds or thousands of volts. That won't damage an ordinary diode at all, since the current is limited. Check the capabilities of any production tester (or an ancient Tek curve tracer).

Mighty Teradyne's ($2.5bn market cap) first product was a production diode tester.

These guys (and their Taiwan competitors) make simple diode production testers (and handlers that run at ~20,000 pieces/hour)

Here's a 40K/hour handler (30ms test time)

Hmm.. one factory (~300,000 square feet) says they can make 8 million diodes per day, so they'd have to have the equivalent of 8 or 9 of those machines. Not a big deal at all. In fact, they'd need that kind of automated handling and test machinery to hope to make any money at a fraction of a cent each. 8E6 at 0.25 cent each is only $20K a day gross (If you were as intelligent as you pretend you would at least realize if

They don't 100% test everything on the data sheet of a complex device, just the important parameters such as PIV and maybe Vf at a couple currents, to weed out the most common issues. In fact many data sheets explicitly state that a few parameters (eg. input capacitance) are guaranteed by characterization rather than 100% testing. There are lots of process control activities going on as well, so lots of opportunities for w*nkers^H^H^H^H professionals with clipboards.

A real engineering?

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

I don't see the fraud, If I followed the thread, he said he checked leakage of caps at overvoltage, and then suggested it would be interesting to test C vs V. It would be. Or is this just an old resentment resurfacing? Mikek

but

l

ld

sues

The reference Stretto gave makes it clear that they do test 100% of their goods.

NT

The reference Stretto gave makes it clear that they do test 100% of their goods.

-----

right......

That?s why if you scroll towards the end they list the types of tests they do and they all talk about "control lots" and sample sizes...

But what do you know about statistics?

Our experience suggests that most diode makers have at least two different wafer processes, one for low voltage diodes and one for higher ones. The HV diodes behave like PIN diodes (which they pretty much are) and have higher forward drops.

All your arguments are based on the premise that reverse breakdown voltage measurements are destructive. They aren't.

--------

Um, are you that innane? If you don't do a real breakdown test then you don't know the real location it breaks down. Do you not realize that if you do not run a real breakdown destruction test then you cannot truly know the breakdown location? There is a difference between an approximation and extrapolation and actually knowing the true location.

Approximations may work but only after you know the characteristic behavior. Any any case I wouldn't want to be using a component that was pushed to it's limits then thrown back in the bin and deemed ok because it didn't "fail". If would be like testing every car by running it into a wall at 20 mph and the ones that still work then being sold.

A simple test can be run on every diode that determines if the diode works in some range that they were designed to but tests the permanently and negatively alter the device would simply not be done to every device in a production run... I take that back, it would be done by retards such as those that run our government... which is why they are in government and not actually producing anything.

"The breakdown voltage of a material is not a definite value because it is a form of failure and there is a statistical probability whether the material will fail at a given voltage. When a value is given it is usually the mean breakdown voltage of a large sample."

Of course I don't expect someone as egotistical as you to admit when your wrong and imagine your response would be somehow to downplay wiki... typical...

Beating, hanging, flailing and otherwise torturing innocent capacitors is a reportable offense! You have been warned by the Society for the Prevention of Cruelty to Capacitors to IMMEDIATELY cease and desist!

On a sunny day (Thu, 16 Jun 2011 14:38:07 -0500) it happened "Stretto" wrote in :

OK, I think I overlooked that comment.

OK, I agree they do take samples. I would think that if the sample test is good, then the machine works OK, and the other ones must be good too. It is a simple mechanical machine, no mysterious variables. Of course in the case of the BC109 without anything in it I had, I had to do fault finding at a high altitude with danger to my life..... found the 'defective' board, took it back to the company, made a test setup, found that transistor. Half a day or more work + travel. The different value from color code resistors I had only took me some hour to find, started looking because things did not work right, Like I said in an other post, I think for mil spec things are (or should be)

100% tested, save for aviation and space (rad tested too), dunno about medical, maybe Joerg knows if he uses special components. Many components (chips for example) have 'not for life support systems' in their data sheet, did you notice that?

On a sunny day (Thu, 16 Jun 2011 15:38:30 -0700) it happened Jim Thompson wrote in :

But Jim, that is the Republican way! And the Democrat way too, so basically the Capitalist way. What you got against that?????

Jim is turning leftist it seems ;-)

Umm, what do you know about diode manufacturing?

John

You've obviously never read a diode data sheet, so let me help:

Look at Fig 4. No diodes were killed to acquire that data.

Do you not realize that if you

Wrong, wrong, wrong. These diodes are specified for 1 uA max current at specified Vr. THAT's what they would test for. Actually, they could apply some small constant current, like that 1 uA maybe, and measure the resulting diode voltage, and guardband that to be somewhat above the guaranteed Vr.

There is a difference between an approximation and

The required leap of faith would be to assume that the diodes won't fail at, say, 400 volts when their tested voltage drop is, say, 500 volts at 1 uA.

There are too many things that could go wrong in fabbing diodes - wafer defects, dicing problems, soldering, bonding maybe, encapsulation - that testing every diode would be the only way to get acceptable quality. If the production failure rate were, say, 0.1%, and you did offline statistics that told you that the rate was 0.1%, you'd have to reject every production lot, and never ship anything. Testing every one lets you ship.

John